ARTICLE

Received 10 Nov 2010 | Accepted 15 Feb 2011 | Published 15 Mar 2011 DOI: 10.1038/ncomms1239 Inflammation driven by tumour-specific Th1 cells protects against B-cell cancer

Ole Audun Werner Haabeth1, Kristina Berg Lorvik1, Clara Hammarström2, Ian M. Donaldson3,4, Guttorm Haraldsen2, Bjarne Bogen1 & Alexandre Corthay1

The immune system can both promote and suppress cancer. Chronic inflammation and proinflammatory cytokines such as interleukin (IL)-1 and IL-6 are considered to betumour promoting. In contrast, the exact nature of protective antitumour immunity remains obscure. Here, we quantify locally secreted cytokines during primary immune responses against myeloma and B-cell lymphoma in mice. Strikingly, successful cancer immunosurveillance mediated by tumour-specific CD4 + T cells is consistently associated with elevated local levels of both proinflammatory (IL-1α, IL-1β and IL-6) and T helper 1 (Th1)-associated cytokines (interferon-γ (IFN-γ), IL-2 and IL-12). Cancer eradication is achieved by a collaboration between tumour- specific Th1 cells and tumour-infiltrating, antigen-presenting macrophages. Th1 cells induce secretion of IL-1β and IL-6 by macrophages. Th1-derived IFN-γ is shown to render macrophages directly cytotoxic to cancer cells, and to induce macrophages to secrete the angiostatic chemokines CXCL9/MIG and CXCL10/IP-10. Thus, inflammation, when driven by tumour- specific Th1 cells, may prevent rather than promote cancer.

1 Centre for Immune Regulation, Institute of Immunology, University of Oslo and Oslo University Hospital Rikshospitalet, PO Box 4950 Nydalen, 0424 Oslo, Norway. 2 Department of Pathology, Institute of Pathology, Oslo University Hospital Rikshospitalet and University of Oslo, PO Box 4950 Nydalen, 0424 Oslo, Norway. 3 The Biotechnology Centre of Oslo, University of Oslo, PO Box 1125 Blindern, 0317 Oslo, Norway. 4 Department of Molecular Biosciences, University of Oslo, PO Box 1041 Blindern, 0316 Oslo, Norway. Correspondence and requests for materials should be addressed to A.C. (email: [email protected]). nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications  © 2011 Macmillan Publishers Limited. All rights reserved. ARTICLE nature communications | DOI: 10.1038/ncomms1239

he immune system can protect against cancer1. Elevated we provide evidence for a dual antitumour role of Th1-derived IFN-γ. numbers of intratumoral T cells predict long-term survival First, IFN-γ triggers tumouricidal activity of tumour-infiltrating for patients with advanced ovarian carcinoma and colorec- macrophages. Second, IFN-γ induces macrophages to secrete T 2,3 tal cancer . Yet, little is known about the exact nature of protec- the angiostatic chemokines CXCL9/MIG (monokine induced by tive antitumour immune responses. On the other hand, it is well IFN-γ) and CXCL10/IP-10 (IFN-γ inducible protein 10), which may established that chronic inflammation predisposes to cancer. halt tumour progression by inhibiting angiogenesis. Collectively, Proinflammatory cytokines such as interleukin (IL)-1 and IL-6 are our data suggest a cancer-protective role of inflammation driven by considered to be essential for tumour progression, and anti-inflam- tumour-specific Th1 cells. matory drugs have been suggested to treat cancer4–9. However, anti-inflammatory treatments may potentially suppress protective Results antitumour immunity. Strategies to fight malignancies should be The Matrigel cytokine assay. We have developed a method to based on stimulating rather than suppressing the ongoing immune quantify locally secreted cytokines during the early stages of an response against cancer. Therefore, it is crucial to better understand immune response against cancer in mice. The method is based on the interplay between immune cells, inflammation and cancer. s.c. inoculation of cancer cells embedded in a Matrigel collagen gel Tumour-specific CD4 + T cells orchestrate the immune response (Fig. 1a). Matrigel is derived from a murine sarcoma and therefore against cancer. CD4 + T cells are required for cytokine-mediated represents a genuine tumour cell microenvironment19. Matrigel, activation of tumour-specific cytotoxic CD8 + T cells, but they can which is liquid at + 4 °C, gelifies at body temperature and forms a also eliminate cancer in the absence of CD8 + T cells10,11. A recent gel plug containing the injected cancer cells18. The gel functions as study of patients with breast cancer concluded that high numbers of an extracellular matrix and contains the cytokines that are secreted CD4 + T cells in lymph nodes (LNs) predict disease-free survival12. locally by cancer cells and by infiltrating immune cells. At various time In lung and liver cancer, high CD4:CD8 T-cell ratios were associated points after injection, the Matrigel plug can be excised and dissolved with good prognosis13,14. However, CD4 + T cells may also suppress in vitro with collagenase. A cell-free supernatant containing secreted antitumour immunity15. To clarify the mechanism of cancer pre- cytokines that are present in the Matrigel extracellular matrix is vention by CD4 + T cells, we have used idiotype (Id)-specific T-cell obtained by centrifugation (Fig. 1a). Cytokines in the supernatant receptor transgenic (TCR-TG) mice, which were made homozygous are quantified by multiplex bead assays. An initial experiment to test for the severe combined immunodeficiency (SCID) mutation to the method revealed that numerous cytokines could be detected in prevent rearrangement of endogenous TCR chains11. In these mice, the Matrigel at day + 8, which represents the height of the antitumour tumour-specific CD4 + T cells recognize an Id peptide from the immune response18 (Supplementary Fig. S1). To identify cytokines variable region of the immunoglobulin light chain of the MOPC315 associated with successful cancer immunosurveillance, we designed myeloma, presented on major histocompatibility complex (MHC) a series of five experimental situations to compare cytokine levels class II molecules16. Id-specific TCR-TG SCID mice are resistant during successful versus failed immune responses against cancer. against subcutaneous (s.c.) inoculation with syngeneic MOPC315 myeloma cells or with Id-transfected F9 B-lymphoma cells, Tumour-specific CD4 + T cells control local cytokine levels. In whereas non-transgenic mice develop fatal tumours. Protection is the first experimental situation (myeloma ± specific T cells), we Id-specific, CD4 + T cell-mediated, and does not require the presence compared tumour-specific CD4 + T cells with CD4 + T cells that of B cells and CD8 + T cells11,17. recognized an irrelevant antigen, ovalbumin (OVA). Id-specific To study the mechanisms of cancer rejection by Id-specific TCR-TG SCID mice, OVA-specific TCR-TG SCID mice and SCID TCR-TG mice, we have developed a strategy consisting of embed- mice were injected s.c. with MOPC315 myeloma cells. Id-specific ding injected tumour cells in a collagen gel (Matrigel). The Matrigel TCR-TG SCID mice were protected against MOPC315 (Fig. 1b), as functions as an extracellular matrix in which infiltrating immune previously reported11. Control SCID mice, which lack T and B cells, cells can be analysed at various time points after injection. Using succumbed from cancer within 40 days. OVA-specific TCR-TG this method, we have reported the first characterization of a suc- SCID mice, which have T cells but cannot recognize Id, developed cessful primary antitumour immune response initiated by naïve tumours as quickly as SCID mice (Fig. 1b). For a cytokine assay, CD4 + T cells18. In brief, we could show that s.c. injected MOPC315 Id-specific and OVA-specific TCR-TG SCID mice were injected with myeloma cells were surrounded within 3 days by macrophages, MOPC315 in Matrigel (Fig. 1c–g). The total numbers of MOPC315 which captured tumour-specific antigens. Within 6 days, naïve cells and CD11b + macrophages in Matrigel at day + 8 were simi- Id-specific CD4 + T cells became activated in draining LN and sub- lar for both groups (Fig. 1d,e). However, when we used MHC class sequently migrated to the incipient tumour site. On recognition of II upregulation as an activation marker for macrophages18, acti- tumour-derived Id peptides presented on MHC class II molecules vated MHC class IIhigh macrophages were only found in Id-specific by macrophages, Id-specific CD4 + T cells were shown to secrete TCR-TG SCID mice (Fig. 1f). Quantification of 33 cytokines in interferon-γ (IFN-γ). Matrigel-infiltrating macrophages became acti­ Matrigel revealed that the local concentration of 21 cytokines was vated by T cell-derived IFN-γ, and could kill MHC class II-nega- significantly higher in the presence of tumour-specific CD4 + T cells, tive myeloma cells directly18. However, in this previous report the that is, in tumour-resistant Id-specific TCR-TG SCID mice Fig.( 1g; exact function of IFN-γ was not fully defined and the involvement Supplementary Fig. S2). Notably, cytokine levels in serum were of other cytokines was not investigated. similar between the groups, underscoring the importance of meas- In this study, we have further developed the Matrigel assay to uring cytokines locally at the incipient tumour site (Supplementary quantify locally secreted cytokines during primary antitumour Fig. S2). immune responses. Using this method, we uncovered a common core of nine cytokines that were consistently associated with success­ Cytokines associated with myeloma immunosurveillance. We ful cancer immunosurveillance. Strikingly, this core includes both then investigated the role of tumour-specific antigen using an anti- proinflammatory (IL-1α, IL-1β and IL-6) and T helper (Th)1-asso- gen-loss (Ag-loss) variant of MOPC315 which is not rejected by Id- ciated (IL-2, IL-3, IL-12, IFN-γ, CXCL9 and CXCL10) cytokines. specific TCR-TG SCID mice20. In the second experimental situation Twelve additional cytokines were associated with cancer prevention (myeloma ± antigen, one flank), two groups of Id-specific TCR-TG in most, but not all experimental settings investigated. Thus, we SCID mice were injected with Matrigel containing either MOPC315 have identified a total of 21 cytokines, which may serve as a basis to or Ag-loss MOPC315 (Fig. 2a,b). Matrigel cytokine quantification develop cytokine-based immunotherapy for cancer. Furthermore, at day + 8 revealed that the local concentration of 21 cytokines was

 nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications © 2011 Macmillan Publishers Limited. All rights reserved. nature communications | DOI: 10.1038/ncomms1239 ARTICLE

a b 100 IL-1 80 30 min Extracellular Matrigel 60 +cancer 37 °C IL-2 cytokines in * cells IFN-γ supernatant 40 Spin 20

Cells in Percent survival Collagenase 0 pellet 0 10 20 30 40 50 Days post MOPC315 injection

c g CCL2 CCL4 MOPC315 MOPC315 CCL5 CXCL9 CXCL10 G-CSF Id-specific OVA-specific Versus GM-CSF TCR-TG SCID TCR-TG SCID IFN-γ IL-1α IL-1β d e f IL-2 IL-3 NS NS 1,000 * IL-5 106 106 IL-6 750 IL-9 4 4 10 IL-12p40 10 500 IL12p70 2 2 10 10 250 IL-15 IL-17

0 0 MHC class II levels MOPC315 per plug IL-18 10 Macrophages per plug 10 0 on macrophages (MFI) TNF-α Id-sp Id-sp Id-sp 0 1 2 3 4 5 6 OVA-sp OVA-sp OVA-sp 10 10 10 10 10 10 10 pg ml–1

Figure 1 | Role of tumour-specific CD4 + T cells for myeloma immunosurveillance. (a) The Matrigel cytokine assay. Cancer cells are mixed with cold liquid Matrigel before s.c. injection. Matrigel gelifiesin vivo and forms a gel plug. The plug contains the injected cancer cells, infiltrating immune cells and cytokines that are secreted locally. At various time points after injection, the Matrigel plug is excised and dissolved with collagenase. Following centrifugation, cytokines in supernatant are quantified by Luminex technology. Cells in the pellet are analysed by various techniques such as flow cytometry, Luminex technology and gene expression microarrays. (b) Tumour challenge experiment. Id-specific TCR-TG SCID mice (squares), OVA- specific TCR-TGS CID mice (triangles) and SCID mice (circles) were injected s.c. with MOPC315 myeloma cells in phosphate-buffered saline (PBS). Tumour development and survival were followed over time. *P = 0.0004 for difference in survival between Id-specific andO VA-specific TCR-TGS CID mice (log-rank test). (c) Design of the first experimental situation for cytokine analysis (myeloma ± specific T cells). Id-specific TCR-TG SCID mice and OVA- specific TCR-TGS CID mice were injected s.c. with MOPC315 myeloma cells in Matrigel. The plugs were analysed individually at day + 8. (d) Total number of MOPC315 cells per plug. (e) Total number of CD11b + macrophages per plug. (f) MHC class II levels on Matrigel-infiltrating CD11b + macrophages. MFI, mean fluorescence intensity. P* = 0.009, Mann–Whitney test. (g) The concentration of 33 cytokines in the extracellular matrix of the Matrigel plugs was quantified for Id-specific (black bars) andO VA-specific (grey bars) TCR-TG SCID mice. Only cytokines with significantly P( < 0.05, Mann–Whitney test) higher levels in one group compared with the other are included in the bar graph. (b) n = 6–8. (c–g) n = 11. Id-sp, Id-specific TCR-TG SCID mice; NS, not significant; OVA-sp, OVA-specific TCR-TG SCID mice. All data are presented as mean ± s.d. G-CSF, granulocyte colony-stimulating factor; GM-CSF, granulocyte-macrophage colony-stimulating factor; TNF-α, tumor necrosis factor α. significantly higher when cancer immunity was successful, that is, confirming the importance of antigen presentation in situ for in mice injected with antigen-producing myeloma cells (Fig. 2b; T cell-mediated activation of macrophages18. Multiplex bead analy- Supplementary Fig. S3). In the third experimental situation sis revealed that local antigen recognition by tumour-specific CD4 + (myeloma ± antigen, two flanks), we designed a system in which T cells resulted in increased levels of 13 cytokines and decreased successful and failed immunosurveillance could be investigated in levels of 3 cytokines in Matrigel (Fig. 2g; Supplementary Fig. S4). the same mouse at different anatomic locations. Id-specific TCR- Analysis of all the MOPC315 data revealed that 11 cytokines were TG SCID or SCID mice were injected with MOPC315 in Matrigel consistently associated with successful immunosurveillance against on the right flank and with Ag-loss MOPC315 in Matrigel on the myeloma (Table 1). left flank. SCID mice developed tumours with similar kinetics on both flanks. In contrast, Id-specific TCR-TG SCID mice efficiently Cytokines consistently associated with cancer prevention. To suppressed the growth of MOPC315, but failed to reject Ag-loss assess the general significance of our findings, we performed an MOPC315 on the contralateral flank, demonstrating the antigen experiment with another cancer type, namely the F9 B-cell lym- specificity of the antitumour immune response within the same phoma. F9 is an A20 BALB/c-derived lymphoma that was trans- mouse (Fig. 2c). For a Matrigel cytokine assay, Id-specific TCR- fected with Id, whereas F55 is an Id-negative control lymphoma21. TG SCID mice were injected on both flanks with either MOPC315 As F9 but not F55 cells are rejected by Id-specific TCR-TG mice17, or Ag-loss MOPC315 (Fig. 2d–g). At day + 8, activated CD69 + this model represents successful versus failed immunosurveil- tumour-specific CD4 + T cells were observed in the right-flank LN lance against lymphoma. An important difference between the draining the MOPC315 injection site, but not in the contralateral two cancer types is that MOPC315 cells do not express MHC class LN draining Ag-loss MOPC315 (Fig. 2e). Activated, MHC class IIhigh II molecules, whereas F9 cells express high levels and thus can macrophages were only detected in the right-flank Matrigel plugs present Id directly to CD4 + T cells21. Id-specific TCR-TG SCID containing MOPC315, but not in the contralateral site (Fig. 2f), mice were injected with F9 cells in Matrigel on the right flank, and nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications  © 2011 Macmillan Publishers Limited. All rights reserved. ARTICLE nature communications | DOI: 10.1038/ncomms1239

a d

MOPC315 Ag-loss Ag-loss Versus MOPC315 MOPC315 MOPC315

b e f 50 * CCL2 40 Right flank LN CCL4 1,000 30 (MOPC315) CXCL1 20 CXCL9 10 750 CXCL10 0 FGF-2 50 500 G-CSF 40 Left flank LN GM-CSF 30 250 20 (Ag-loss) IFN-γ 10 macrophages (MFI) 0 IL-1α MHC class II levels on Counts 0 IL-1 0 1 2 3 4 Right flank Left flank β 10 10 10 10 10 IL-2 (MOPC315) (Ag-loss) IL-3 CD69 IL-5 IL-6 IL-9 g CCL4 IL-12p70 IL-15 CXCL9 IL-17 CXCL10 IL-18 IFN-γ TNF-α IL-1α 100 101 102 103 104 105 106 IL-1β IL-2 pg ml–1 IL-3 IL-6 c SCID TCR-TG SCID IL-9 20 IL-10 15 IL-12p40 IL-12p70 10 0 1 2 3 4 5 6 Tumour 5 10 10 10 10 10 10 10

diameter (mm) 0 CXCL2 0 5 10 15 20 0 5 10 15 20 M-CSF Days post injection TGF-β

100 101 102 103 104 105 106 pg ml–1

Figure 2 | Role of antigen for myeloma immunosurveillance. (a) Design of the second experimental situation for cytokine analysis (myeloma ± antigen, 1 flank). Id-specific TCR-TG SCID mice were injected s.c. with MOPC315 or with Ag-loss MOPC315 in Matrigel. At day + 8, the concentration of 33 cytokines in the extracellular matrix of the Matrigel plugs was quantified for each mouse. b( ) Matrigel cytokine levels for mice inoculated with MOPC315 (black bars) or with Ag-loss MOPC315 (grey bars) from the experiment described in a. Only cytokines with significantly P( < 0.05) higher levels in one group compared with the other are included in the bar graph. (c) Double-flank tumour challenge experiment. Id-specific TCR-TG SCID or SCID mice were injected with MOPC315 in Matrigel on the right flank (black diamonds), and with Ag-lossMO PC315 in Matrigel on the left flank (grey circles). Tumour development on each side was followed over time. (d) Design of the third experimental situation for cytokine analysis (myeloma ± antigen, 2 flanks). Id-specific TCR-TG SCID mice were injected with MOPC315 on the right flank, and with Ag-loss MOPC315 on the left flank.M atrigel plugs and draining LNs were analysed at day + 8. (e) Expression of CD69 on gated tumour-specific (GB113 + ), CD4 + T cells in the LN draining the right and left flank from a representative mouse. Dotted lines indicate an isotype-matched control monoclonal antibodies (mAb). (f) MHC class II levels on gated CD11b + macrophages that had infiltrated the Matrigel plugs containing either MOPC315 (right flank) or Ag-loss MOPC315 (left flank). P* = 0.0002. (g) Cytokines levels in Matrigel plugs containing either MOPC315 (black bars) or Ag-loss MOPC315 (grey bars). Only cytokines with significantly higher (top) or lower (bottom) levels (P < 0.05) in one group compared with the other are included in the graphs. (a, b) n = 6. (c) n = 7–8. (d–g) n = 11. All data are presented as mean ± s.d. Statistical analysis was performed with the Mann–Whitney test. FGF-2, fibroblast growth factor 2;M -CSF, macrophage colony-stimulating factor; TGF-β, transforming growth factor-β.

with F55 cells in Matrigel on the left flank. In addition, SCID mice with left-flank Matrigel plugs containing F55Fig. ( 3c; Supplementary were injected with F9 cells in Matrigel (Fig. 3a–c). Activated, MHC Fig. S5). In a second parallel analysis of the data (lymphoma ± specific class IIhigh macrophages were only detected in Matrigel containing Id- T cells), Matrigel plugs containing F9 were compared between TCR- expressing F9 in TCR-TG SCID mice, but not in Matrigel with F55, TG SCID and SCID mice, revealing increased levels of 20 cytokines nor in SCID mice (Fig. 3b). Thus, the activation of Matrigel-infil- in the situation in which immunosurveillance against lymphoma trating macrophages required (i) the presence of Id-specific CD4 + was successful, that is, in Id-specific TCR-TG mice Table( 1; Supple- T cells and (ii) the local production of Id antigen. This observation mentary Fig. S5). Analysis of all the F9 data showed that 16 cytokines suggests that the antitumour immune response may be very simi- were consistently associated with successful immunosurveillance lar against MOPC315 and F9 cells, although the two cell types differ against B-cell lymphoma (Table 1). Further analysis revealed a com- in terms of MHC class II expression. Cytokine quantification of all mon core of nine cytokines that were consistently associated with suc- Matrigel plugs allowed two separate analyses of the data. First, cytokine cessful cancer immunosurveillance against myeloma and lymphoma: levels were compared between each flank of the TCR-TG SCID mice IL-1α, IL-1β, IL-2, IL-3, IL-6, IL-12p70, IFN-γ, CXCL9 and CXCL10 (lymphoma ± antigen, two flanks). The local levels of 19 cytokines (Table 1). Twelve additional cytokines were associated with cancer were higher in right-flank Matrigel plugs containing F9, compared rejection in most but not all situations investigated (Table 1).

 nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications © 2011 Macmillan Publishers Limited. All rights reserved. nature communications | DOI: 10.1038/ncomms1239 ARTICLE

Table 1 | Cytokine profiles of successful cancer immunosurveillance.*

Experiment Experiment Experiment Experiment Experiment Cytokines Cytokines myeloma ± specific myeloma ± antigen myeloma ± antigen lymphoma ± antigen lymphoma ± specific always often T cells 1 flank 2 flanks 2 flanks T cells present† present‡ IL-1α IL-1α IL-1α IL-1α IL-1α IL-1α IL-1β IL-1β IL-1β IL-1β IL-1β IL-1β IL-2 IL-2 IL-2 IL-2 IL-2 IL-2 IL-3 IL-3 IL-3 IL-3 IL-3 IL-3 IL-6 IL-6 IL-6 IL-6 IL-6 IL-6 IL-12p70 IL-12p70 IL-12p70 IL-12p70 IL-12p70 IL-12p70 IFN-γ IFN-γ IFN-γ IFN-γ IFN-γ IFN-γ CXCL9 CXCL9 CXCL9 CXCL9 CXCL9 CXCL9 CXCL10 CXCL10 CXCL10 CXCL10 CXCL10 CXCL10 CCL4 CCL4 CCL4 CCL4 IL-9 IL-9 IL-9 IL-9 CCL2 CCL2 CCL2 CCL2 CCL2 CCL5 CCL5 CCL5 CCL5 FGF-2 FGF-2 FGF-2 FGF-2 G-CSF G-CSF G-CSF G-CSF G-CSF IL-5 IL-5 IL-5 IL-5 IL-5 IL-12p40 IL-12p40 IL-12p40 IL-12p40 IL-12p40 LIF§ LIF CXCL1 CXCL1 GM-CSF GM-CSF GM-CSF GM-CSF IL-10 IL-15 IL-15 IL-15 IL-15 IL-17 IL-17 IL-18 IL-18 IL-18 IL-18 M-CSF M-CSF TNF-α TNF-α TNF-α TNF-α *In five different experimental situations, locally secreted cytokines were quantified during successful versus failed immunity against myeloma and B-cell lymphoma. For each situation, cytokines with higher levels (P < 0.05) in the group with successful cancer immunosurveillance are listed. †Cytokines always associated with successful immunosurveillance, that is, in five out of five situations tested. ‡Cytokines often but not always associated with successful immunosurveillance, that is, in 3–4 out of 5 situations tested. §LIF, leukemia inhibitory factor.

Role of IFN-. IFN-γ-deficient (IFN-γ − / − ) Id-specific TCR-TG SCID T cells, because macrophages isolated from SCID mice, even in high mice were unable to reject MOPC315 or F9 cells, demonstrating the numbers, could not inhibit MOPC315 growth. It was speculated critical role of IFN-γ for immunosurveillance of both myeloma and that macrophages were rendered tumouricidal by Th1-derived IFN- lymphoma (Fig. 4a). Interestingly, tumour growth was significantly γ18. To test this hypothesis, we compared the tumouricidal activity delayed in IFN-γ − / − Id-specific TCR-TG SCID mice as compared of macrophages isolated from either IFN-γ + / + or IFN-γ − / − TCR-TG with SCID mice, suggesting residual, but insufficient antitumour SCID mice. The ability of Matrigel-infiltrating macrophages from immunity in the absence of IFN-γ (Fig. 4a). For a Matrigel cytokine IFN-γ − / − mice to inhibit the growth of MOPC315 cells in vitro assay, IFN-γ + / + and IFN-γ − / − Id-specific TCR-TG SCID mice were was severely reduced, although not completely abolished (Fig. 4g). injected with MOPC315 (Fig. 4b–f). At day + 8, activated CD69 + Hence, IFN-γ is required for rendering tumour-infiltrating macro- Id-specific T cells were found for both groups of mice in draining phages fully competent to directly kill cancer cells. LNs and in Matrigel plugs (Fig. 4c). Thus, IFN-γ was required nei- ther for activation of naïve tumour-specific T cells in LNs nor for Myeloma immunosurveillance is mediated by Th1 cells. IFN-γ is a their migration and reactivation at the incipient tumour site. In typical product of activated Th1 cells but tumour-specific Th17 cells fact, significantly more tumour-specific T cells were found in the may also secrete IFN-γ 23. Intracellular flow cytometry revealed that Matrigel plugs from IFN-γ − / − mice (Fig. 4d), presumably because Matrigel-infiltrating Id-specific T cells produced IFN-γ and tumour T-cell proliferation is generally stronger in IFN-γ − / − mice22. Although necrosis factor α (TNF-α), and expressed the transcription factor activated Matrigel-infiltrating tumour-specific T cells were present T-bet, which is consistent with a Th1 differentiation. In contrast, in both groups, activated MHC class IIhigh macrophages were only Id-specific T cells did not express molecular markers for Th2 (IL-4, detected in the plugs from IFN-γ + / + mice (Fig. 4e), confirming the Gata-3), Th17 (IL-17) or regulatory T cells (Foxp3; Fig. 5a). Thus, importance of IFN-γ for macrophage activation18. Matrigel cytokine myeloma rejection in Id-specific TCR-TG SCID mice is mediated quantification revealed that the local levels of three cytokines by IFN-γ-producing tumour-specific Th1 cells. (IFN-γ, CXCL9 and CXCL10) were significantly lower, whereas one cytokine (IL-1β) was higher, in IFN-γ − / − compared with IFN- Cytokine production of tumouricidal macrophages. We then γ + / + mice (Fig. 4f; Supplementary Fig. S6). Thus, IFN-γ deficiency quantified intracellular cytokines in sorted populations of Matrigel- results in reduced secretion of CXCL9 and CXCL10 at the incipient infiltrating macrophages. Macrophages involved in tumour rejec- tumour site. tion were purified from Id-specific TCR-TG SCID mice injected In a previous report, we have shown that Matrigel-infiltrating with MOPC315, whereas control macrophages were from SCID macrophages, isolated from Id-specific TCR-TG SCID mice, were mice. Successful antitumour immunity was associated with elevated cytotoxic to MOPC315 cells in vitro18. Tumouricidal macrophage levels of nine cytokines in Matrigel-infiltrating macrophages (Fig. activity was strictly dependent on macrophage activation by CD4 + 5b; Supplementary Fig. S7). These data suggest that, among the nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications  © 2011 Macmillan Publishers Limited. All rights reserved. ARTICLE nature communications | DOI: 10.1038/ncomms1239

a b * Id-specific TCR-TG SCID mice was characterized by the absence * of angiogenesis, whereas formation of CD31 + blood vessels around ) 1,500 tumour islets was readily observable in control mice (Fig. 5g,h).

1,000 This suggests that IFN-γ-activated macrophages use simultaneously F55 F9 F9 two strategies to fight cancer: direct killing of cancer cells Fig.( 4g) 500 and indirect inhibition of tumour growth through secretion of angiostatic chemokines such as CXCL9 and CXCL10 (Fig. 5b–h). MHC class II levels

TCR-TG SCID SCID on macrophages (MFI 0 F9 F55 F9 Myeloma immunosurveillance is an inflammatory reaction. His- TCR-TG SCID tological analysis revealed that myeloma rejection by Id-specific c TCR-TG SCID mice was characterized by local inflammation with CCL2 oedema and inflammatory cell infiltrate (Fig. 6a). Matrigel-infiltrat- CCL5 ing macrophages were further characterized by gene expression CXCL1 microarrays (Fig. 6b) and flow cytometry (Fig. 6c). Tumouricidal CXCL9 macrophages were isolated from Id-specific TCR-TG SCID mice CXCL10 injected with MOPC315-containing Matrigel. Control macro- FGF-2 phages were obtained from TCR-TG SCID mice injected with G-CSF Matrigel containing Ag-loss MOPC315. The expression of CXCL9 IFN-γ IL-1α and CXCL10 was upregulated in tumouricidal macrophages com- IL-1β pared with controls, confirming that these angiostatic chemokines IL-2 were produced by T cell-activated macrophages (Fig. 6b, left panel). IL-3 Interestingly, several other chemokines were upregulated, includ- IL-5 ing CXCL11/I-TAC (interferon-inducible T-cell α chemokine), IL-6 another IFN-γ-inducible angiostatic chemokine, CCL17/TARC IL-12p40 (thymus and activation regulated chemokine) and CCL22/MDC IL-12p70 (macrophage-derived chemokine). The levels of CCL17 and CCL22 IL-15 in Matrigel samples from the myeloma experiments were measured LIF by enzyme-linked immunosorbent assay, but no association with M-CSF successful immunosurveillance was found (Supplementary Fig. S8). 100 101 102 103 104 105 106 Tumouricidal macrophages showed upregulation of genes associ- pg ml–1 ated with antigen processing and presentation (Fig. 6b, right panel), and increased levels of MHC class II and co-stimulatory molecules Figure 3 | Immunosurveillance of B-cell lymphoma. (a) Design of the such as CD40, CD80, CD86, CD273 and CD274 (Fig. 6c). Hence, fourth (lymphoma ± antigen, 2 flanks) and fifth (lymphoma ± specific tumouricidal macrophages have a typical phenotype of activated T cells) experimental situations for cytokine analysis. Id-specific TCR-TG professional antigen-presenting cells. SCID mice were injected s.c. with Id + F9 B-lymphoma cells in Matrigel on Gene expression analysis revealed a number of cytotoxicity-asso- the right flank, and with Id − F55 B-lymphoma cells in Matrigel on the left ciated genes, which were upregulated in tumouricidal macrophages flank.S CID mice were injected s.c. with Id + F9 cells in Matrigel on the right and which may be involved in the direct killing of cancer cells: pro- flank.M atrigel plugs were analysed at day + 8. (b) Expression of MHC teases (mcpt8, mmp12); granzyme A and B; ass1 and cybb (involved class II on gated CD11b + macrophages that had infiltrated the Matrigel in nitric oxide and superoxide production, respectively); and several plugs for each group of samples. (c) Cytokines in Matrigel were quantified. genes known to be expressed by natural killer (NK) cells (CD161/ Data for Id + F9 (black bars) versus Id − F55 (grey bars) in TCR-TG SCID klrb1b, CD314/NKG2D, Ly49b/klra2 and rnf19b; Fig. 6b, middle mice are shown in the bar graph for cytokines with significantly P( < 0.05) panel; Supplementary Data S1 and S2). Moderate upregulation of higher levels in one group compared with the other. (a–c) n = 10–11. NKG2D was confirmed at protein level (Fig. 6c). The presence of *P < 0.001. All data are presented as mean ± s.d. Statistical analysis was NK-associated genes in the microarray data is unlikely to originate performed with the Mann–Whitney test. from cell contamination because we could not detect Matrigel-infil- trating NKp46hi NK cells by flow cytometry. Furthermore, we per- formed an experiment with beige mutant mice with a functional defect in NK-cell lytic functions24. Id-specific TCR-TG SCID beige cytokines consistently associated with successful cancer immuno- mice were fully protected against challenge with MOPC315, indicat- surveillance, CXCL9, CXCL10, IL-1β and IL-6, were produced by ing that NK cells were not involved in myeloma rejection (Supple- macrophages on activation by tumour-specific CD4 + T cells. Two mentary Fig. S9). Bioinformatic analysis of the upregulated genes in additional cytokines (IL-1α and IL-12p70) are also likely to be tumouricidal macrophages showed highly significant overlaps with macrophage-derived because they were present in macrophages, genes or proteins known to be induced in macrophages by IFN-γ although their levels were not significantly different between (Supplementary Data S3). Furthermore, promoter region analysis TCR-TG SCID and SCID mice (Supplementary Fig. S7). The three revealed that the upregulated genes were enriched for binding sites remaining cytokines in the common core (IL-2, IL-3 and IFN-γ) for inflammation-associated transcription factors such as nuclear are typical T-cell products, and therefore most likely secreted by factor-κB and RelA (Supplementary Data S3). Thus, myeloma Matrigel-infiltrating Id-specific Th1 cells. immunosurveillance is an inflammatory reaction characterized by Production of CXCL10 by macrophages was further investigated tissue infiltration by IFN-γ-activated macrophages with a typical by immunohistochemistry. CXCL10 was detected in F4/80 + mac- inflammatory phenotype. rophages that penetrated the Matrigel plugs containing CD138 + MOPC315 cells (Fig. 5c,d). Matrigel-infiltrating macrophages Discussion produced more CXCL10 in tumour-resistant Id-specific TCR-TG We have developed a method to quantify locally secreted cytokines SCID mice as compared with tumour-prone OVA-specific mice during primary antitumour immune responses in mice. Using (Fig. 5e,f). Strikingly, successful myeloma immunosurveillance in this method, we have identified a common core of nine cytokines

 nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications © 2011 Macmillan Publishers Limited. All rights reserved. nature communications | DOI: 10.1038/ncomms1239 ARTICLE

a 100 100

80 80 ** 60 60

40 *** 40 Percent survival

Percent survival 20 20 ***

0 * 0 0 10 20 30 40 50 60 0 10 20 30 40 50 60 70 80 Days post MOPC315 myeloma injection Days post F9 B-lymphoma injection

c d e b 20 +/+ *** 15 IFN-γ * 750 10 106 MOPC315 MOPC315 5 500 Versus 0 4 40 10 –/– 30 IFN-γ 2 250 20 10

10 T cells per plug +/+ –/– MHC class II levels Counts 0 IFN-γ IFN-γ 0 10 on macrophages (MFI) 0 TCR-TG SCID TCR-TG SCID 100 101 102 103 +/+ –/– +/+ –/– CD69

f g 175 ) CXCL9 –3 150 10 × CXCL10 125

IFN-γ 100 75 101 102 103 104 105 106 50

25 IL-1β MOPC315 growth (cpm 0 101 102 103 104 105 106 MOPC 20:1 10:1 4:1 1:1 315 –1 Macrophages + MOPC315 pg ml only (effector/target ratio)

Figure 4 | Role of IFN-γ for tumour protection and local cytokine production. (a) Tumour challenge experiments. IFN-γ + / + Id-specific TCR-TG SCID mice (black squares), IFN-γ − / − Id-specific TCR-TGS CID mice (grey squares) and IFN-γ + / + SCID mice (dotted line) were injected s.c. with MOPC315 myeloma cells (left panel) or with F9 B-lymphoma cells (right panel) in PBS. Tumour development and survival were followed over time by palpation. (b) Experimental design for Matrigel cytokine assay. Id-specific TCR-TG SCID IFN-γ + / + and IFN-γ − / − mice were injected s.c. with MOPC315 in Matrigel. At day + 8, Matrigel-infiltrating T cells and macrophages were analysed, and locally secreted cytokines were quantified.c ( ) Expression of CD69 on Matrigel-infiltrating, tumour-specific T cells (GB113 + TCRCβ + ) from representative IFN-γ + / + and IFN-γ − / − mice. Dotted lines indicate an isotype-matched control mAb. (d) Total number of tumour-specific T cells (GB113 + TCRCβ + ) per plug for IFN-γ + / + and IFN-γ − / − mice. (e) Expression of MHC class II on gated CD11b + macrophages that had infiltrated the Matrigel plugs in IFN-γ + / + and IFN-γ − / − mice. (f) The concentration of 33 cytokines in Matrigel was quantified.O nly cytokines with significantly higher (top) or lower (bottom) levels P( < 0.05) in IFN-γ + / + (black bars) mice compared with IFN-γ − / − (grey bars) mice are included in the bar graphs. (g) Growth inhibition assay. Id-specific TCR-TG SCID IFN-γ + / + and IFN-γ − / − mice were injected s.c. with Matrigel containing MOPC315. Matrigel-infiltrating CD11b + macrophages were purified at day + 7 and tested at various effector to target ratios for their ability to suppress the proliferation of MOPC315 cells in vitro. Black bars, macrophages from IFN-γ+/+ mice; grey bars, macrophages from IFN-γ−/− mice. (a) n = 5–14. (b–f) n = 11. (g) triplicates. *P < 0.05, **P < 0.01, ***P < 0.001. Statistical analysis was performed with the log-rank test (a) and the Mann–Whitney test (d–f). All data are presented as mean ± s.d. consistently associated with successful cancer eradication: IL-1α, associated with shorter survival and disease progression in patients IL-1β, IL-2, IL-3, IL-6, IL-12p70, IFN-γ, CXCL9 and CXCL10. To with chronic myelogenous leukaemia and colorectal carcinoma5,27. our knowledge, these data represent the first characterization of the In mice, IL-1 was shown to be required for tumour invasiveness and cytokine profile of successful cancer immunosurveillance. Interest- angiogenesis4,28. In several human malignancies, including multiple ingly, in patients with hepatocellular carcinoma, increased expres- myeloma, lymphoma and lung cancer, high serum levels of IL-6 were sion of 17 genes in the liver, including IL-1α, IL-1β, IL-2, IL-12 and associated with poor prognosis29–31. Furthermore, IL-6 was shown IFN-γ, was associated with good prognosis25. Thus, the cytokine to be required for obesity-induced tumour promotion in a mouse profile of protective immunity against cancer may be very similar model for hepatocellular carcinoma9. Although inflammation is for mice and humans. The observation that IL-12 and IFN-γ were generally considered to be tumour promoting, a few histopathologi- constantly associated with cancer rejection is consistent with a Th1 cal studies in colorectal, breast and bladder cancer indicated that polarization of the immune response, which is widely considered to tumour infiltration by inflammatory cells may be associated with be favourable for immunological control of cancer1–3,26. In contrast, better prognosis32–34. To reconcile these opposing views, we propose the presence in the list of the proinflammatory cytokines IL-1α, that inflammation, when driven by tumour-specific Th1 cells, may IL-1β and IL-6 may seem more surprising. Chronic inflammation prevent cancer (Fig. 7a). In a Th1 environment, proinflammatory predisposes to cancer and cancer-related inflammation is generally cytokines (IL-1α, IL-1β and IL-6) may participate in cancer eradica- regarded as being tumour promoting4–9. Elevated IL-1β levels were tion by recruiting leucocytes from the circulation and by stimulating nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications  © 2011 Macmillan Publishers Limited. All rights reserved. ARTICLE nature communications | DOI: 10.1038/ncomms1239

a b 103 102 CCL2 101 CCL5 GB113 100 CXCL9 0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 CXCL10 IgG1 IFN-� TNF-� IL-4 IL-17 IL-1β 103 IL-6 102 IL-9

1

GB113 10 IL-12p40

100 TNF-α 100 101 102 103 100 101 102 103 100 101 102 103 100 101 102 103 100 101 102 103 IgG1 T-bet IgG2b Gata-3 pg ml–1 103

102

101 GB113 100 100 101 102 103 100 101 102 103 IgG2a Foxp3 e c d M M M M M M M M M

F4/80 F4/80 CXCL10 CD138 CXCL10 CD138 Hoechst F4/80 CXCL10 Hoechst Hoechst f g h

M M M M M M M M

CXCL10 CD31 CD31 CD138 CD138 CD138 Hoechst Hoechst Hoechst

Figure 5 | Cytokine production of tumour-specific CD4 + T cells and macrophages. (a) Characterization of Matrigel-infiltrating tumour-specific T cells by intracellular flow cytometry. Id-specific TCR-TGS CID mice (n = 12–15) were injected s.c. with MOPC315 in Matrigel. At day + 8, Matrigel plugs were excised and pooled. Matrigel-infiltrating tumour-specific (GB113 + TCRCβ + ) T cells were stained for intracellular cytokines and for transcription factors. IgG1, IgG2a and IgG2b indicate isotype-matched control mAb. (b) Measurement of intracellular cytokines in Matrigel-infiltrating macrophages. Id-specific TCR-TG SCID mice or SCID mice (n = 12–15) were injected with MOPC315 in Matrigel. At day + 8, Matrigel-infiltrating CD11b + macrophages were purified and pooled for groups of 3–4 mice. Sorted macrophages were washed and lysed. Cytokines in cell lysates were quantified by Luminex technology. The bar graph represents cytokine levels in macrophages from Id-specific TCR-TG SCID mice (black bars) and from SCID mice (grey bars) for cytokines with significantly P( < 0.05, Mann–Whitney test) higher levels in one group compared with the other (mean ± s.d.). For the purpose of comparison, cytokine levels were corrected to 2×105 macrophages per pool. (c) Visualization of successful myeloma immunosurveillance. MOPC315 myeloma cells were injected in Matrigel into a tumour-resistant Id-specific TCR-TG SCID mouse. At day + 8, a tissue sample containing the Matrigel plug was excised for immunohistochemistry. Matrigel-infiltrating macrophages were stained with anti-F4/80 (green) mAb.MO PC315 myeloma cells were stained with anti-CD138 (red) mAb. Hoechst stain (blue) was used to visualize cell nuclei. (d) A day + 8, sample from a tumour-resistant Id-specific TCR-TG SCID mouse was double-stained with F4/80 (red) and CXCL10 (green). (e) CXCL10 (green) production by Matrigel-infiltrating macrophages in a tumour-resistant Id-specific TCR-TGS CID mouse. MOPC315 myeloma cells were stained with anti-CD138 mAb (red). (f) CXCL10 (green) production by Matrigel-infiltrating macrophages in a tumour-proneO VA-specific TCR-TG SCID mouse. (g) Anti-CD31 polyclonal antibodies were used to stain blood vessels in a tumour-resistant Id-specific TCR-TG SCID mouse. (h) Formation of CD31 + (green) blood vessels around CD138 + (red) myeloma tumour islets in a tumour-prone SCID mouse. Immunostaining images (c–h) show representative day + 8 specimens for groups of ≥3 mice. Scale bars, 50 µm. M, edge of Matrigel plug.

 nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications © 2011 Macmillan Publishers Limited. All rights reserved. nature communications | DOI: 10.1038/ncomms1239 ARTICLE

a

D Inf

Muscle M Muscle M M M M T T M T T T T

b MHC Protein MHC MHC I Cd1d cxcl9 mcpt8 antigen II-� II-� Protein ccl17 Granzyme B antigen Immunoproteaso cxcl11 ass1 Endocytosis Psmb8 ccl5 Granzyme A Psme MHC 1

MHC MHC I 9 ccl22 CD161/klrb1b II- II-

� � me 10 2 cxcl10 CD314/NKG2D MHC MHC DM-� DM-� ccl24 mmp12 Peptide Peptide loading fragments CLIP cxcl13 Ly49b/klra2 Late MHC MHC Tap1 endosome II-� II-� cxcl7 rnf19b Tap2 ER cxcl14 cybb Peptide loading Tapasin MHC log2 MHC MHC MHC II-� II-� I I –6 –3 0 +3 +6

c 40 MHC class II CD40 CD80 CD86 CD273 CD274 NKG2D 30

s 20 10

Count 0

Fluorescence intensity

Figure 6 | Myeloma immunosurveillance is an inflammatory reaction. (a) Tumour-resistant Id-specific and tumour-prone OVA-specific TCR-TGS CID mice were injected s.c. with MOPC315 myeloma cells in Matrigel. Tissue samples were taken at day + 8 and sections were stained with hematoxylin and eosin. Successful myeloma immunosurveillance in an Id-specific TCR-TGS CID mouse is characterized by inflammation with oedema and infiltration of inflammatory cells (left panel). Such inflammatory features are not observed in a control OVA-specific TCR-TGS CID mouse (right panel). Representative images for groups of ≥3 mice are shown. Scale bars, 100 µm. D, dermis; Inf, inflammatory cell infiltrate;M , edge of Matrigel plug; T, tumour cell islet. (b) Gene expression was compared for tumouricidal/control macrophages. Matrigel-infiltrating, tumouricidal macrophages were isolated at day + 8 from Id-specific TCR-TG SCID mice (n = 12–15) injected with MOPC315-containing Matrigel. Control macrophages were obtained from Id-specific TCR- TG SCID mice (n = 12–15) injected with Matrigel containing Ag-loss MOPC315. The expression of all known genes was quantified by gene microarrays. Data for chemokines (left panel) and cytotoxicity-associated genes (middle panel) that were differentially expressed are shown. Each column indicates one independent experiment. The cartoon (right panel) illustrates the antigen processing and presentation pathway. Proteins encoded by genes whose expression was consistently ≥2 times higher in tumouricidal macrophages in three independent experiments are shown in yellow. ass1, argininosuccinate synthetase 1; cybb, cytochrome b-245 beta polypeptide; ER, endoplasmic reticulum; mcpt8, mast cell protease 8; mmp12, matrix metalloproteinase 12; Psm, proteasome subunit; rnf19b, ring finger protein 19B; Tap, transporter associated with antigen processing. c( ) Characterization of Matrigel-infiltrating macrophages at day + 8 by flow cytometry. GatedM atrigel-infiltrating CD11b + cells were analysed in Id-specific TCR-TGS CID mice injected s.c. with either MOPC315 (shaded blue area) or Ag-loss MOPC315 (boldface line). Dotted lines indicate isotype-matched control mAb. Multiple experiments were conducted with similar results.

CD4 + T-cell functions35,36. In contrast, other types of inflammation, tion of acute-phase response and fever. The IL-1 receptor and the or inflammation that lacks tumour specificity, may not protect and Toll-like receptors share a common intracellular signalling path- may even promote tumour development (Fig. 7b). Hence, caution way, which leads to activation of the nuclear factor-κB and to the should be taken when considering anti-inflammatory treatments transcription of several proinflammatory cytokines, including IL-1 against cancer4–7. Drugs that may dampen the tumour-suppressive itself. Thus, IL-1 functions as a positive feedback loop in inflamma- Th1-driven inflammatory immune response should preferably be tion. IL-1 was shown to enhance the expansion and differentiation avoided. of CD4 + T cells35,36, and to induce macrophage tumouricidal activity The finding that elevated levels of IL-1α ( and β) were consist- in vitro37. Importantly, IL-1β secretion by macrophages is dependent ently associated with successful cancer immunosurveillance is of on activation of the inflammasome, a cytosolic molecular complex particular interest. IL-1 is a canonical proinflammatory cytokine, responsible for generating active IL-1β by cleaving the inactive pre- which exerts a broad range of activities including activation of cursor. The inflammasome functions as a guardian of the body by vascular endothelium, stimulation of lymphocyte functions, induc- detecting pathogens and danger38. In cancer immunosurveillance, nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications  © 2011 Macmillan Publishers Limited. All rights reserved. ARTICLE nature communications | DOI: 10.1038/ncomms1239

CXCL9 Angiogenesis cytokines to treat patients with cancer. Significant benefits have IFN-γ CXCL10 Recruitment of Th1 been demonstrated in the treatment of various malignancies with 53–56 IL-2 and CTL cells TNF-α, IL-2 and IFN-α . Other candidate molecules such as granulocyte-macrophage colony-stimulating factor, IL-7, IL-12p70 Th1 M 57,58 Φ Direct killing and IL-15 have shown promising antitumour effects . How- IL-3 of cancer cells ever, only a few cytokines are being routinely used to treat cancer. IL-12 Vascular A major challenge to develop cytokine-based immunotherapy IL-1α 59 permeability is that cytokine networks are based on synergistic interactions . IL-1β Therefore, coadministration of different cytokines is likely to be IL-6 T-cell required to efficiently treat cancer, as suggested by experiments functions with mice59,60. With more than 200 different cytokines described to date, the number of combinations to be tested for cancer therapy is MΦ enormous. Our study identified 9 cytokines that were consistently Angiogenesis associated with successful antitumour immunity, and 12 additional cytokines that were associated with cancer eradication in most Vascular IL-1α but not all situations investigated. This list should be helpful for permeability IL-1β the development of combined cytokine-based immunotherapy for IL-6 Tumour cancer. invasiveness Methods Figure 7 | A model for how inflammation may either suppress or Mice, cell lines and injection of tumour cells. Heterozygous Id-specific TCR-TG 11 − / − promote cancer. (a) Tumour-suppressive inflammation.S uccessful SCID mice or SCID littermates were on BALB/c background . IFN-γ BALB/c mice22 (from The Jackson Laboratory) were crossed with Id-specific TCR-TG SCID cancer immunosurveillance is mediated by a tumour-specific Th1-driven mice to obtain IFN-γ − / − Id-specific TCR-TG SCID mice. OVA-specific TCR- inflammation. In this process, tumour-specific Th1 cells collaborate with TG SCID mice were generated by introducing the scid mutation into DO11.10 tumour-infiltrating macrophages, and nine cytokines synergize to prevent TCR-TG mice (from Taconic). MOPC315 (IgA, λ2315) is a transplantable BALB/c cancer. Proinflammatory cytokines (IL-1α, IL-1β and IL-6) participate in plasmacytoma obtained from the American Type Culture Collection (ATCC) and tumour eradication by recruiting immune cells from the circulation and by propagated as in vitro growing cells. Ag-loss MOPC315.36 (ref. 20) was kindly provided by Alexander Marks, University of Toronto. F9 is an A20 BALB/c-derived stimulating T-cell functions. (b) Tumour-promoting inflammation. In the B-lymphoma cell line that was transfected with Id-containing λ2315 (ref. 21). absence of sufficient numbers of tumour-specific Th1 cells, IL-1α, IL-1β and F55 is an Id-negative lymphoma obtained by transfecting A20 cells with pSV2neo IL-6 may participate in tumour progression by stimulating angiogenesis, vector without insert21. Adult mice were injected s.c., in the interscapular region vascular permeability and tumour invasiveness. MΦ, macrophage; CTL, or in the flank, with 1–1.6×105 cancer cells suspended either in 100 µl phosphate- cytotoxic CD8 + T cells. buffered saline (PBS, Gibco), or in 250 µl Growth Factor-Reduced Matrigel (BD Biosciences). Tumour growth was followed over time by palpation. Mice with tumour a diameter ≥10 mm were killed. The study was approved by the National Committee for Animal Experiments (Oslo, Norway). the nature of the endogenous danger signals detected by the inflam- Matrigel cytokine assay. Cancer cells were mixed with ice-cold Matrigel and masome remains to be determined, but a role for ATP released by 250 µl of the mixture was carefully injected s.c. into each mouse under anaesthesia. 39 Matrigel, which is liquid at + 4 °C, gelifies at body temperature and forms a gel necrotic tumour cells has been suggested . plug containing the injected cancer cells. At day + 8, mice were sacrificed. The s.c. The importance of IFN-γ for cancer immunosurveillance is well Matrigel plug was excised and treated with 1 ml RPMI 1640 medium (Gibco) sup- documented1,40,41, but its exact function remains poorly understood. plemented with 1 mg ml − 1 collagenase type IV from Clostridium histolyticum and IFN-γ was originally defined as a cytokine able to render macro- 0.3 mg ml − 1 DNase I from bovine pancreas (Sigma), at 37 °C for 30 min. Dissolved phages tumouricidal in vitro42. However, the in vivo relevance of Matrigel solution was squeezed through a stainless steel sieve (Sigma) and cen- trifuged 300 g for 7 min. The cell pellet was analysed by flow cytometry. Cell-free this finding remained unclear. Our data with IFN-γ-deficient mice supernatant was pressed through a 0.45 µm syringe filter (PALL Corporation) and provide direct evidence for a role of IFN-γ in inducing macrophage kept at − 70 °C until analysis by Luminex technology. To test the method, tumour- tumouricidal activity in vivo. Furthermore, we demonstrate another resistant Id-specific TCR-TG SCID mice were injected s.c. with either Matrigel function of IFN-γ, namely induction of secretion of MIG/CXCL9 containing MOPC315 cells or with empty Matrigel, as a control. At day + 8, the and IP-10/CXCL10 by tumour-infiltrating macrophages. CXCL10 Matrigel plugs were excised for cytokine analysis. For comparison, a Matrigel 43 solution was used to generate in vitro plugs, which were processed similarly. The selectively attracts Th1 cells and may thus function as a positive concentration of 33 cytokines in the extracellular matrix of the Matrigel plugs was feedback loop in the Th1-driven antitumour immune response. Th1- quantified by Luminex technology. This experiment showed that the Matrigel solu- induced CXCL9 and CXCL10 may also be important for recruiting tion itself (in vitro or ex vivo) only contains very low levels of cytokines, except for CD8 + T cells to the incipient tumour site44. Moreover, neovascu- transforming growth factor-β and vascular endothelial growth factor, which were larization is required for a solid tumour to grow45. Several endog- high. In contrast, numerous cytokines were detected in the day + 8 MOPC315- containing Matrigel plugs recovered from s.c. injections (Supplementary Fig. S1). enous angiostatic factors have been described including CXCL9 and CXCL10, which both binds to the CXCR3-B receptor on endothelial Preparation of macrophage lysates. TCR-TG SCID mice (n = 12) or SCID mice cells, resulting in suppression of endothelial cell proliferation and (n = 14) were injected s.c. with 105 MOPC315 cells in 250 µl Matrigel. At day + 8, differentiation into capillary structures46–48. Treatment with recom- Matrigel-infiltrating CD11b + macrophages were sorted by FACSAria (≥ 95% pure) 5 binant CXCL9 and CXCL10 was shown to delay tumour growth in and pooled for groups of three to four mice. Sorted macrophages (1.5–3×10 per 49–51 pool) were washed and lysed in a buffer containing 5 mM EDTA, 50 mM Tris-HCL mice by inducing vascular damage and tumour necrosis . It has pH7.0, 0.5% NP-40 and 150 mM NaCl (all from Sigma). Lysates were pressed + been previously suggested that tumour-specific CD4 T cells may through a 0.45 µm syringe filter and kept at − 70 °C until analysis by Luminex mediate tumour rejection by inhibiting angiogenesis through secre- technology. tion of IFN-γ 41, but the mechanism remained obscure. Our results provide direct evidence that IFN-γ, secreted locally by tumour- Cytokine quantification by Luminex technology. Cytokine levels in Matrigel specific Th1 cells, induces tumour-infiltrating macrophages to supernatants, cell lysates and serum were measured using single-plex or multiplex bead assays from Bio-Rad Laboratories (Bio-Plex 23-plex and 9-plex panels) and secrete the angiostatic chemokines CXCL9 and CXCL10. Biosource Invitrogen (for TGF-β and CXCL-10), according to the manufacturer’s Since the discovery of the antitumour properties of TNF-α instructions. Samples were analysed as singlets and standards in duplicates, using a in 1975 (ref. 52), there has been a considerable interest in using Luminex-100 instrument with Bio-Plex Manager 4.1 software (Bio-Rad Laboratories).

10 nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications © 2011 Macmillan Publishers Limited. All rights reserved. nature communications | DOI: 10.1038/ncomms1239 ARTICLE

Analysis of cells by flow cytometry. Single-cell suspension from draining axillary 2. Zhang, L. et al. Intratumoral T cells, recurrence, and survival in epithelial LNs and Matrigel plugs were obtained by use of a stainless steel sieve. Unspecific ovarian cancer. N. Engl. J. Med. 348, 203–213 (2003). binding was blocked by incubation with heat-inactivated (56 °C, 30 min) 30% 3. Galon, J. et al. Type, density, and location of immune cells within human normal rat serum in PBS and 100 µg ml − 1 anti-FcγRII/III monoclonal antibodies colorectal tumors predict clinical outcome. Science 313, 1960–1964 (2006). (mAb; clone 2.4G2). Cells were stained for 15 min on ice with specific mAbs in PBS 4. Voronov, E. et al. IL-1 is required for tumor invasiveness and angiogenesis. supplemented with 0.5% bovine serum albumin (Biotest). The following commer- Proc. Natl Acad. Sci. USA 100, 2645–2650 (2003). cially available mAbs were used, conjugated with either fluorescein, phycoerythrin, 5. Konishi, N. et al. Interleukin-1 receptor antagonist inhibits the expression allophycocyanin or biotin: CD4 (RM4.5 or GK1.5), CD11b (M1/70), CD273 of vascular endothelial growth factor in colorectal carcinoma. Oncology 68, (TY25), CD274 (MIH5), CD314 (CX5), IFN-γ (XMG1.2), IL-4 (11B11), IL-17 138–145 (2005). (TC11-18H10), MHC class II I-A/I-E (M5/114.15.2), TCRCβ (H57-597), TNF-α 6. Lin, W. W. & Karin, M. A cytokine-mediated link between innate immunity, (MP6-XT22) (BD Biosciences); CD11b (3A33), CD40 (1C10), CD69 (H1.2F3), inflammation, and cancer.J. Clin. Invest. 117, 1175–1183 (2007). CD80 (1G10), CD86 (GL1) (Southern Biotechnology); Foxp3 (FJK-16s), Gata-3 7. Mantovani, A., Allavena, P., Sica, A. & Balkwill, F. Cancer-related inflammation. (TWAJ), T-bet (4B10) (eBioscience). The following mAbs were affinity-purified Nature 454, 436–444 (2008). and, if needed, biotinylated in our laboratory: anti-Id-specific-TCR-clonotype 8. Kim, S. et al. Carcinoma-produced factors activate myeloid cells through TLR2 (GB113), anti-FcγRII/III (2.4G2; ATCC). Biotinylated mAbs were detected with to stimulate metastasis. Nature 457, 102–106 (2009). streptavidin conjugated to peridinin chlorophyll protein (BD Biosciences). For 9. Park, E. J. et al. Dietary and genetic obesity promote liver inflammation and intracellular cytokine detection, cells were stimulated with phorbol myristate tumorigenesis by enhancing IL-6 and TNF expression. Cell 140, 197–208 acetate and ionomycin (both from Sigma) in vitro for 4 h in cell culture medium (2010). supplemented with monensin, before staining with Cytofix/Cytoperm Plus reagents 10. Levitsky, H. I., Lazenby, A., Hayashi, R. J. & Pardoll, D. M. In vivo priming of (BD Biosciences) and specific mAbs. Quadruple-stained cells were analysed on a two distinct antitumor effector populations: the role of MHC class I expression. FACSCalibur instrument with CellquestPro software (BD Biosciences). J. Exp. Med. 179, 1215–1224 (1994). 11. Bogen, B. et al. Naive CD4+ T cells confer idiotype-specific tumor resistance in Tumour cell growth inhibition assay. Matrigel-infiltrating CD11b + cells were the absence of antibodies. Eur. J. Immunol. 25, 3079–3086 (1995). purified by FACSAria at day + 7 after s.c. injection. Sorted CD11b + cells (≥95% 12. Kohrt, H. E. et al. Profile of immune cells in axillary lymph nodes predicts pure) were irradiated (2,000 rad) and added at various effector/target ratios to MOPC315 disease-free survival in breast cancer. PLoS Med. 2, e284 (2005). cell cultures (104 tumour cells per well) in triplicates. Cultures were pulsed with 13. Nakamura, H. et al. Immunologic parameters as significant prognostic factors [3H]thymidine after 48 h and collected 12 h later on a TopCount NXT microplate counter. in lung cancer. Lung Cancer 37, 161–169 (2002). 14. Unitt, E. et al. Tumour lymphocytic infiltrate and recurrence of hepatocellular Tissue preparation. The Matrigel plugs were removed by a wide excision of the carcinoma following liver transplantation. J. Hepatol. 45, 246–253 (2006). flank wall, including the skin. Tissue samples were either flash frozen in O.C.T. 15. Knutson, K. L., Disis, M. L. & Salazar, L. G. CD4 regulatory T cells in human compound (Tissue-Tek) in liquid nitrogen and stored at − 80 °C, or fixed in 10% cancer pathogenesis. Cancer Immunol. Immunother. 56, 271–285 (2007). buffered formaldehyde and paraffin-embedded. Four micrometre tissue sections 16. Bogen, B., Malissen, B. & Haas, W. Idiotope-specific T cell clones that recognize were cut from the tissue blocks. Frozen sections were air-dried overnight and syngeneic immunoglobulin fragments in the context of class II molecules. subsequently acetone-fixed. Paraffin-embedded sections were deparaffinized with Eur. J. Immunol. 16, 1373–1378 (1986). xylene, and bathed consecutively in absolute ethanol, 96% ethanol, 70% ethanol 17. Lauritzsen, G. F., Weiss, S., Dembic, Z. & Bogen, B. Naive idiotype-specific and PBS. Sections for light microscopy were stained with hematoxylin for 3 min, CD4+ T cells and immunosurveillance of B-cell tumors. Proc. Natl Acad. Sci. hexamine for 3 min and eosin for 45 s, and subsequently dehydrated with xylene. USA 91, 5700–5704 (1994). 18. Corthay, A. et al. Primary antitumor immune response mediated by CD4+ Immunohistochemistry. The paired immunostainings for F4/80 + CD138 and T cells. Immunity 22, 371–383 (2005). F4/80 + CXCL10 were performed on frozen sections, whereas the other stainings 19. Kleinman, H. K. et al. Basement membrane complexes with biological activity. were done on formalin-fixed paraffin-embedded tissue sections. Frozen tissue Biochemistry 25, 312–318 (1986). sections were incubated with the primary and secondary antibodies for 1 h each 20. Corthay, A., Lundin, K. U., Lorvik, K. B., Hofgaard, P. O. & Bogen, B. at room temperature. To visualize CXCL10, we amplified the signal by means of a Secretion of tumor-specific antigen by myeloma cells is required for cancer secondary antibody horseradish peroxidase-conjugated goat anti-rabbit IgG and immunosurveillance by CD4+ T cells. Cancer Res. 69, 5901–5907 (2009). Alexa 488-conjugated tyramide amplification reagent according to manufacturer’s 21. Weiss, S. & Bogen, B. MHC class II-restricted presentation of intracellular instructions (Invitrogen). For paraffin-embedded sections, antigen retrieval was antigen. Cell 64, 767–776 (1991). performed on deparaffinized, formalin-fixed tissue sections by boiling those for 22. Dalton, D. K. et al. Multiple defects of immune cell function in mice with 20 min in 10 mM Tris-EDTA pH 9. Sections were then incubated with the primary disrupted interferon-gamma genes. Science 259, 1739–1742 (1993). antibodies overnight at 4 °C, followed by the secondary antibodies for 1 h at room 23. Muranski, P. et al. Tumor-specific Th17-polarized cells eradicate large temperature. Hoechst blue nuclear stain was added to the final volume of washing established melanoma. Blood 112, 362–373 (2008). buffer after the last incubation. As negative controls, primary irrelevant isotype- 24. Roder, J. & Duwe, A. The beige mutation in the mouse selectively impairs and concentration-matched antibodies were used. Samples were examined with natural killer cell function. Nature 278, 451–453 (1979). a Nikon Eclipse E-800 fluorescence microscope (Nikon) equipped with Nikon 25. Budhu, A. et al. Prediction of venous metastases, recurrence, and prognosis in Plan-Fluor objective lenses and an F-VIEW digital camera controlled by AnalySIS hepatocellular carcinoma based on a unique immune response signature of the 3.2 software (Soft Imaging System GmbH). Antibodies and all working concentra- liver microenvironment. Cancer Cell 10, 99–111 (2006). tions are listed in Supplementary Tables S1 and S2. 26. Finak, G. et al. Stromal gene expression predicts clinical outcome in breast cancer. Nat. Med. 14, 518–527 (2008). Gene expression analysis. Id-specific TCR-TG SCID mice were injected s.c. 27. Wetzler, M. et al. Altered levels of interleukin-1 beta and interleukin-1 receptor with Matrigel containing 105 MOPC315 cells. At day + 8, Matrigel-infiltrating antagonist in chronic myelogenous leukemia: clinical and prognostic correlates. CD11b + tumouricical macrophages were sorted by FACSAria (≥95% pure) in Blood 84, 3142–3147 (1994). three independent experiments. Control macrophages were obtained similarly 28. Vidal-Vanaclocha, F., Amezaga, C., Asumendi, A., Kaplanski, G. & Dinarello, in two independent experiments from Id-specific TCR-TG SCID mice injected C. A. Interleukin-1 receptor blockade reduces the number and size of murine with Matrigel containing Ag-loss MOPC315 cells. TRIzol (Invitrogen) was used B16 melanoma hepatic metastases. Cancer Res. 54, 2667–2672 (1994). to extract mRNA from sorted macrophages. The expression of all known genes 29. Pelliniemi, T. T. et al. Immunoreactive interleukin-6 and acute phase proteins was quantified by gene microarrays (GeneChip Mouse Genome 430 2.0 Array, as prognostic factors in multiple myeloma. Finnish Leukemia Group. Blood 85, Affymetrix). Changes in gene expression ≥( 2) that were consistent for all sample 765–771 (1995). comparisons (tumouricidal/control macrophages) were considered to be signifi- 30. Preti, H. A. et al. Prognostic value of serum interleukin-6 in diffuse large-cell cant. The microarray expression data sets were deposited in Gene Expression lymphoma. Ann. Intern. Med. 127, 186–194 (1997). Omnibus (GEO, www.ncbi.nlm.nih.gov/geo) under accession number GSE26912. 31. Martin, F. et al. Cytokine levels (IL-6 and IFN-gamma), acute phase response and nutritional status as prognostic factors in lung cancer. Cytokine 11, 80–86 Statistical analysis. For tumour challenge experiments, differences in survival (1999). were analysed with the log-rank test. Flow cytometry data and cytokine levels were 32. Klintrup, K. et al. Inflammation and prognosis in colorectal cancer. analysed with the Mann–Whitney test. Statistical analysis was performed with Eur. J. Cancer 41, 2645–2654 (2005). GraphPad Prism 4 software. 33. Lee, A. H. et al. Different patterns of inflammation and prognosis in invasive carcinoma of the breast. Histopathology 48, 692–701 (2006). 34. Cai, T. et al. Prognostic role of the tumor-associated tissue inflammatory References reaction in transitional bladder cell carcinoma. Oncol. Rep. 16, 329–334 (2006). 1. Shankaran, V. et al. IFNgamma and lymphocytes prevent primary tumour 35. Von Stebut, E. et al. Interleukin 1alpha promotes Th1 differentiation and development and shape tumour immunogenicity. Nature 410, 1107–1111 inhibits disease progression in Leishmania major-susceptible BALB/c mice. (2001). J. Exp. Med. 198, 191–199 (2003). nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications 11 © 2011 Macmillan Publishers Limited. All rights reserved. ARTICLE nature communications | DOI: 10.1038/ncomms1239

36. Ben-Sasson, S. Z. et al. IL-1 acts directly on CD4 T cells to enhance their 55. Kirkwood, J. M. et al. Interferon alfa-2b adjuvant therapy of high-risk resected antigen-driven expansion and differentiation.Proc. Natl Acad. Sci. USA 106, cutaneous melanoma: the Eastern Cooperative Oncology Group Trial EST 7119–7124 (2009). 1684. J. Clin. Oncol. 14, 7–17 (1996). 37. Philip, R. & Epstein, L. B. Tumour necrosis factor as immunomodulator and 56. Rohatiner, A. Z. et al. Meta-analysis to evaluate the role of interferon in mediator of monocyte cytotoxicity induced by itself, gamma-interferon and follicular lymphoma. J. Clin. Oncol. 23, 2215–2223 (2005). interleukin-1. Nature 323, 86–89 (1986). 57. Spitler, L. E. et al. Adjuvant therapy of stage III and IV malignant melanoma 38. Martinon, F., Mayor, A. & Tschopp, J. The inflammasomes: guardians of the using granulocyte-macrophage colony-stimulating factor. J. Clin. Oncol. 18, body. Annu. Rev. Immunol. 27, 229–265 (2009). 1614–1621 (2000). 39. Ghiringhelli, F. et al. Activation of the NLRP3 inflammasome in dendritic cells 58. Cheever, M. A. Twelve immunotherapy drugs that could cure cancers. induces IL-1beta-dependent adaptive immunity against tumors. Nat. Med. 15, Immunol. Rev. 222, 357–368 (2008). 1170–1178 (2009). 59. Fleischmann, W. R. Jr, Kleyn, K. M. & Baron, S. Potentiation of antitumor effect 40. Dighe, A. S., Richards, E., Old, L. J. & Schreiber, R. D. Enhanced in vivo growth of virus-induced interferon by mouse immune interferon preparations. J. Natl and resistance to rejection of tumor cells expressing dominant negative IFN Cancer Inst. 65, 963–966 (1980). gamma receptors. Immunity 1, 447–456 (1994). 60. Rosenberg, S. A., Schwarz, S. L. & Spiess, P. J. Combination immunotherapy for 41. Qin, Z. & Blankenstein, T. CD4+ T cell—mediated tumor rejection involves cancer: synergistic antitumor interactions of interleukin-2, alfa interferon, and inhibition of angiogenesis that is dependent on IFN gamma receptor tumor-infiltrating lymphocytes.J. Natl Cancer Inst. 80, 1393–1397 (1988). expression by nonhematopoietic cells. Immunity 12, 677–686 (2000). 42. Schreiber, R. D., Pace, J. L., Russell, S. W., Altman, A. & Katz, D. H. Acknowledgments Macrophage-activating factor produced by a T cell hybridoma: physiochemical We thank Lise Aagaard, Hela Soltani and Siv Haugen Tunheim for technical help; Anita and biosynthetic resemblance to gamma-interferon. J. Immunol. 131, 826–832 Berge, Merete Høyem, Marianne Johannessen, Nina Kjølen, Maureen Lysaker, Thayaruby (1983). Narayanapillai, Mette Nielsen, Christine Olsen and Anneli Skogstad for helping type 43. Bonecchi, R. et al. Differential expression of chemokine receptors and and for taking good care of the animals; Inger Øynebråten, Ludvig A. Munthe, Peter chemotactic responsiveness of type 1 T helper cells (Th1s) and Th2s. O. Hofgaard and Einar Rødland for critical reading of the manuscript. This work was J. Exp. Med. 187, 129–134 (1998). supported by grants from Anders Jahre fund, Henrik Homans Minde fund, Legat for 44. Nakanishi, Y., Lu, B., Gerard, C. & Iwasaki, A. CD8(+) T lymphocyte fremme av kreftforskningen, Familien Blix fond til fremme for medisinsk forskning, mobilization to virus-infected tissue requires CD4(+) T-cell help. Nature 462, S.G. Sønneland Foundation, Andrine og Hans Gysler Berg fund, the Multiple Myeloma 510–513 (2009). Research Foundation, Helse Sør-Øst and the Research Council of Norway. 45. Folkman, J. Tumor angiogenesis: therapeutic implications. N. Engl. J. Med. 285, 1182–1186 (1971). 46. Luster, A. D., Greenberg, S. M. & Leder, P. The IP-10 chemokine binds to Author contributions a specific cell surface heparan sulfate site shared with platelet factor 4 and O.A.W.H., K.B.L., C.H., I.M.D., G.H. and A.C. performed the experiments, collected the inhibits endothelial cell proliferation. J. Exp. Med. 182, 219–231 (1995). data and prepared the figures. All authors analysed and discussed the data. A.C. and B.B. 47. Angiolillo, A. L. et al. Human interferon-inducible protein 10 is a potent designed the study. A.C. wrote the manuscript. O.A.W.H., K.B.L., C.H., I.M.D. and B.B. inhibitor of angiogenesis in vivo. J. Exp. Med. 182, 155–162 (1995). contributed in writing the manuscript. K.B.L. and C.H. contributed equally to this work. 48. Lasagni, L. et al. An alternatively spliced variant of CXCR3 mediates the inhibition of endothelial cell growth induced by IP-10, Mig, and I-TAC, and Additional information acts as functional receptor for platelet factor 4. J. Exp. Med. 197, 1537–1549 Accession codes: Expression data sets were deposited in GEO under accession number (2003). GSE26912. 49. Sgadari, C. et al. Interferon-inducible protein-10 identified as a mediator of tumor necrosis in vivo. Proc. Natl Acad. Sci. USA 93, 13791–13796 (1996). Supplementary Information accompanies this paper at http://www.nature.com/ 50. Arenberg, D. A. et al. Interferon-gamma-inducible protein 10 (IP-10) is an naturecommunications angiostatic factor that inhibits human non-small cell lung cancer (NSCLC) tumorigenesis and spontaneous metastases. J. Exp. Med. 184, 981–992 (1996). Competing financial interests: The authors declare no competing financial interests. 51. Sgadari, C. et al. Mig, the monokine induced by interferon-gamma, promotes tumor necrosis in vivo. Blood 89, 2635–2643 (1997). Reprints and permission information is available online at http://npg.nature.com/ 52. Carswell, E. A. et al. An endotoxin-induced serum factor that causes necrosis of reprintsandpermissions/ tumors. Proc. Natl Acad. Sci. USA 72, 3666–3670 (1975). How to cite this article: Haabeth, O.A.W. et al. Inflammation driven by tumour-specific 53. Cherix, S. et al. Isolated limb perfusion with tumor necrosis factor and melphalan Th1 cells protects against B-cell cancer.Nat. Commun. 2:240 doi: 10.1038/ncomms1239 for non-resectable soft tissue sarcomas: long-term results on efficacy and limb (2011). salvage in a selected group of patients. J. Surg. Oncol. 98, 148–155 (2008). 54. Fyfe, G. et al. Results of treatment of 255 patients with metastatic renal cell License: This work is licensed under a Creative Commons Attribution-NonCommercial- carcinoma who received high-dose recombinant interleukin-2 therapy. Share Alike 3.0 Unported License. To view a copy of this license, visit http:// J. Clin. Oncol. 13, 688–696 (1995). creativecommons.org/licenses/by-nc-sa/3.0/

12 nature communications | 2:240 | DOI: 10.1038/ncomms1239 | www.nature.com/naturecommunications © 2011 Macmillan Publishers Limited. All rights reserved.